20260903T1532Z prd e6 e7 m1

What. Added E6 (groundedness of identifying claims, P0) and E7 (typed partials, v0 kinds person | organization, per-kind identifying class, P0). Extended M1 with alignment stability under re-extraction and identifying-claim error rate per source type, computed against a labeled eval set. Milestone 3 done-when now requires the eval-set plan before the measured corpus is extracted; milestone 5 notes organizations-first bootstrap (D26).

Why. D22, D23, D25, D26; rationale in docs/design/v0-review.md sections A and B.

Verify. docs/PRD.md section 6 (E6, E7 rows; M1 row) and section 7 (milestones 3, 5).